Re: pgcrypto compilation error due to stack-allocated EVP_CIPHER_CTX

On Mon, Dec 5, 2016 at 6:09 PM, Michael Paquier
<michael(dot)paquier(at)gmail(dot)com> wrote:
> On Mon, Dec 5, 2016 at 5:11 PM, Heikki Linnakangas <hlinnaka(at)iki(dot)fi> wrote:
>> I'm afraid if we just start using EVP_CIPHER_CTX_new(), we'll leak the
>> context on any error. We had exactly the same problem with EVP_MD_CTX_init
>> being removed, in the patch that added OpenSSL 1.1.0 support. We'll have to
>> use a resource owner to track it, just like we did with EVP_MD_CTX in commit
>> 593d4e47. Want to do that, or should I?
>
> I'll send a patch within 24 hours.

And within the delay, attached is the promised patch.

While testing with a linked list, I have found out that the linked
list could leak with cases like that, where decryption and encryption
are done in a single transaction;
select pgp_sym_decrypt(pgp_sym_encrypt(repeat('x',65530),'1'),'1') =
repeat('x',65530);

What has taken me a bit of time was to figure out that this bit is
needed to free correctly elements in the open list:
@@ -219,6 +220,8 @@ encrypt_free(void *priv)
{
struct EncStat *st = priv;

+ if (st->ciph)
+ pgp_cfb_free(st->ciph);
px_memset(st, 0, sizeof(*st));
px_free(st);
}
This does not matter on back-branches as things get cleaned up once
the transaction memory context gets freed.
--
Michael